Software Engineer, Systems ML - Frameworks / Compilers / DL-Kernels
Meta PlatformsJob Title
Software Engineer, Systems ML - Frameworks / Compilers / DL-Kernels
Role Summary
Join the MTIA (Meta Training & Inference Accelerator) Software team within AI and Compute Foundations to develop software that enables high-performance training and inference on specialized AI hardware. The role focuses on PyTorch/GenAI frameworks, AI compilers and runtimes, or high-performance deep learning kernels and tooling to optimize ML workloads on current and next-generation MTIA platforms.
Experience Level
Mid-level. Preferred experience varies by credential: typical guidance in the posting maps to seniority by degree (see Education Requirements for details).
Responsibilities
Primary responsibilities include platform and performance work across framework, compiler, runtime, and kernel domains.
- Develop software for one of these core focus areas: AI frameworks, compiler stack, high-performance DL kernels, runtime, or tooling for MTIA accelerators.
- Contribute to PyTorch and GenAI framework components and core compiler development to support new inference and training accelerators.
- Analyze deep learning models and implement optimization algorithms to improve performance and efficiency.
- Collaborate with AI researchers and hardware teams to co-design and deploy optimizations for ML models on MTIA hardware.
- Apply software development best practices for feature design, optimization, and performance tuning.
Requirements
Must-have technical skills and relevant experience; preferred items are noted separately.
- Must-have: Proven C/C++ programming skills.
- Must-have: Experience in AI framework development or accelerating deep learning models on hardware architectures.
- Nice-to-have: Experience with frameworks and runtimes such as PyTorch, TensorFlow, ONNX, TensorRT, or GenAI frameworks (vLLM, SGLang).
- Nice-to-have: Compiler optimization experience (loop optimizations, vectorization, parallelization, SIMD) and familiarity with MLIR, LLVM, IREE, XLA, TVM, or Halide.
- Nice-to-have: Runtime and system performance optimization experience (latency, memory bandwidth, I/O, compute utilization) and associated tooling.
- Nice-to-have: High-performance kernel development experience (CUDA, OpenMP, OpenCL) or accelerator kernel programming; familiarity with libraries like cuBLAS, cuDNN, CUTLASS, HIP/ROCm.
- Nice-to-have: Tooling experience for device profiling, tracing, debugging, and performance tuning.
- Nice-to-have: Knowledge of GPU, CPU, or AI hardware accelerator architectures.
- Nice-to-have: Demonstrated ongoing AI skill development and experience implementing responsible/ethical AI practices (risk assessment, bias mitigation, quality reviews).
Education Requirements
Posting references a Bachelor's degree in Computer Science, Computer Engineering, or a relevant technical field (or equivalent practical experience). Preferred mappings: Bachelor's + 7+ years, Master's + 4+ years, or PhD + 3+ years in AI framework development or accelerating deep learning models on hardware architectures.
About the Company
Company: Meta Platforms
Headquarters: Menlo Park, California, United States
American technology company that develops social networking products (Facebook, Instagram, WhatsApp) and invests in virtual/augmented reality hardware and software through Reality Labs, focusing on connectivity, advertising, and immersive computing experiences.
